---
license: apache-2.0
datasets:
- alecsharpie/nailbiting_classification
language:
- en
base_model:
- google/siglip2-base-patch16-224
pipeline_tag: image-classification
library_name: transformers
tags:
- Nailbiting
- Human
- Behaviour
- siglip2
---

![NB.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/rziUroDd0QVnPpXbys6zv.png)

# **NailbitingNet**

> **NailbitingNet** is a binary image classification model based on `google/siglip2-base-patch16-224`, designed to detect **nail-biting behavior** in images. Leveraging the **SiglipForImageClassification** architecture, this model is ideal for behavior monitoring, wellness applications, and human activity recognition.

```py
Classification Report:
              precision    recall  f1-score   support

      biting     0.8412    0.9076    0.8731      2824
   no biting     0.9271    0.8728    0.8991      3805

    accuracy                         0.8876      6629
   macro avg     0.8841    0.8902    0.8861      6629
weighted avg     0.8905    0.8876    0.8881      6629
```

![download.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/SW6xjZzA7eViFsAmrmxdR.png)

---

## **Label Classes**

The model distinguishes between:

```
Class 0: "biting"         → The person appears to be biting their nails  
Class 1: "no biting"      → No nail-biting behavior detected
```

---

## **Installation**

```bash
pip install transformers torch pillow gradio
```

---

## **Example Inference Code**

```python
import gradio as gr
from transformers import AutoImageProcessor, SiglipForImageClassification
from PIL import Image
import torch

# Load model and processor
model_name = "prithivMLmods/NailbitingNet"
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# ID to label mapping
id2label = {
    "0": "biting",
    "1": "no biting"
}

def detect_nailbiting(image):
    image = Image.fromarray(image).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()

    prediction = {id2label[str(i)]: round(probs[i], 3) for i in range(len(probs))}
    return prediction

# Gradio Interface
iface = gr.Interface(
    fn=detect_nailbiting,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(num_top_classes=2, label="Nail-Biting Detection"),
    title="NailbitingNet",
    description="Upload an image to classify whether the person is biting their nails or not."
)

if __name__ == "__main__":
    iface.launch()
```

---

## **Use Cases**

* **Wellness & Habit Monitoring**
* **Behavioral AI Applications**
* **Mental Health Tools**
* **Dataset Filtering for Behavior Recognition**